红外甲烷传感器温度校正模型研究  

Research on Temperature Rectification Model of IR Methane Detector

作　　者：张帆[1] 张立萍[2] 魏丽[1] 王天杰[1] 赵晨光[1] 刘志刚[1] 

机构地区：[1]唐山学院,河北唐山063000 [2]河北能源职业技术学院,河北唐山063004

出　　处：《煤矿机械》2009年第10期43-45,共3页Coal Mine Machinery

基　　金：唐山市科技攻关项目(08160202A-2-2)

摘　　要：采用BP神经网络,建立了红外甲烷传感器温度校正模型。以隐层节点数为变量,研究了神经网络规模对训练速度和推广能力的影响。将贝叶斯正规化方法与Levenberg-Marquardt优化算法相结合,提高了神经网络训练的效率和推广能力。所建立的模型网络规模小,运算速度快,可以对温度变化引起的红外甲烷传感器非线性误差进行自动校正。BP neural network has been applied to the development of the temperature rectification model of IR methane detector. The number of neurons in hidden layer is adopted as an index to the size of the network, whose influence on the learning speed and generalization capacity of the neural network has been thoroughly investigated. Bayesian regularization in combination with Levenberg- Marquardt algorithm has been applied to achieving faster learning speed and well-generalized neural network. The obtained neural network is small in size and exhibits fast computing speed. Therefore, the non-linear error of the IR methane detector in respect of temperature deviation can be automatically rectified.

关 键 词：红外甲烷传感器 BP神经网络 温度校正 贝叶斯正规化 

分 类 号：TD77[矿业工程—矿井通风与安全] TP183[自动化与计算机技术—控制理论与控制工程]